Cray was awarded a contract back in June 2013 to provide the European Centre for Medium-Range Weather Forecasts (ECMWF) with a Cray Sonexion storage system and two Cray XC30 supercomputers. A $36 million deal was signed now by Cray for upgrading its supercomputers used by ECMWF.
The European weather prediction center provides up to 15 days medium-range forecasts of global weather as well as seasonal and monthly forecasts. The supercomputer system is having its headquarters in Reading, England.
The system contains the world's largest archive of numerical weather prediction data and is one of the largest in the world of meteorology. Among the high performance calculations performed by the Cray weather forecast supercomputer is included a sophisticated global prediction model of the oceans and atmosphere.
The planned upgrade will make possible developing the complex models of the research center for providing even more accurate weather predictions. Cray will upgrade its supercomputers at ECMWF under the terms of the new deal. During this year Cray will deliver the new systems to the weather research facility.
The old Cray XC30 systems will be replaced with new Cray XC40 systems, according to a report published by Scientific Computing. ECMWF will also expand its research facility with a 32-node Cray XC40-AC system with the next-generation of the Knights Landing Intel Xeon Phi processor and additional Cray Sonexion 2000 scale-out Lustre storage.
Intel unveiled officially the Knights Landing 14-nanometer processor in November 2015, at the SC 15 supercomputer show. The Intel chip features as many as 72 core that allow parallel processing and can play the role of either a co-processor or primary processor. The Knights Landing chip is using Intel Silvermont architecture and contains over 8 billion transistors. According to Intel, the processor can reach over 3 teraflops of peak performance.
According to ZD Net, citing Erland Källén, ECMWF Director of Research, the upgrade will allow the European weather forecast center to develop high-resolution predictions of severe weather events up to 15 days ahead.
